CUTS DISTRIBUTION COSTS 18% LOGISTICS PYTHON

Network Optimisation Model

Python linear programming model (PuLP + scipy) evaluating warehouse locations, lane allocations, and transport mode assignments to minimise total distribution cost while satisfying service level constraints across a 12 node European network.

Project Overview

⚑ Problem

Suboptimal Legacy Network

The distribution network was designed 8 years ago for a different customer mix. Four warehouse locations were chosen based on real estate availability, not demand centroids. Demand has shifted significantly since, resulting in 23% of lanes operating well above optimal cost per km and 14% of SLA commitments at risk due to distance misalignment.

⚙ Solution

LP Network Optimisation (PuLP)

A Python model formulates the network as a minimum cost flow problem: minimise Σ(cost per lane × volume) subject to supply balance at each warehouse, demand satisfaction at each customer node, capacity constraints, and maximum service lead time per SLA band. Candidate warehouse locations evaluated via p median heuristic.

✓ Findings

18% Cost Reduction €1.4M Saved

The optimised network consolidates from 4 to 3 active DCs, shifts 8 road lanes to rail, and eliminates 6 suboptimal cross docking points. Total annual distribution cost drops from €7.8M to €6.4M an 18% saving. SLA compliance improves from 81% to 97%. Full implementation over 9 months.

Interactive Network Visualisation

European Distribution Network Current vs Optimised

Warehouse / DC

Customer Node

Closed (Optimised)

Road

Rail

Sea

Eliminated Lane

CURRENT TOTAL COST

€7.82M

Annual distribution

OPTIMISED TOTAL COST

€6.41M

▼ 18.0% reduction

ANNUAL SAVING

€1.41M

Vs current network

SLA COMPLIANCE

97%

▲ from 81% current

DCs ACTIVE

3 / 4

1 DC consolidated

LANES SHIFTED TO RAIL

Road → Rail modal shift

Dataset Explorer

Network Cost Analysis

Current vs Optimised Cost per Corridor (€K)

Mode Split Current vs Optimised (Volume)

Distance vs Cost/Unit Bubble by Volume

SLA Compliance by Corridor

Savings Opportunity by Lane (€K)

Cost per Unit KM by Mode

Lane Register

LANE ID	ORIGIN	DESTINATION	MODE	DIST (KM)	VOLUME (UNITS)	TRANSPORT (€)	HANDLING (€)	TOTAL COST (€)	OPT SAVING (€)	SLA DAYS	ACTUAL DAYS	SLA MET	STATUS

Cost Reduction Roadmap

Monthly Distribution Cost Current to Optimised (12 Month Implementation)

Company Advice

💡 Recommendations for Logistics Strategy Teams

Network redesign typically delivers 15 25% cost reduction but only if the model constraints are realistic. Include maximum service lead time per customer tier and warehouse capacity ceilings as hard constraints, not soft guidelines.
Run the model annually at minimum: demand pattern shifts, fuel costs, and new customer acquisitions can erode an optimised network back to suboptimal within 18 months without re validation.
Before closing any DC, model the transition cost (inventory rebalancing, lease break penalties, customer communication) against the steady state saving. Payback under 18 months is the typical go/no go threshold.
Rail modal shift only pencils out above 400km lane distance. Below this threshold, the handling cost premium and longer lead time erode the freight saving validate your minimum distance before applying the model recommendation.
Build in a network resilience constraint: the LP will always consolidate to the theoretical minimum nodes, but single node failure risk increases proportionally. Require at minimum dual source coverage for all Tier 1 customer zones.
Involve commercial and customer service teams early a network that cuts cost but degrades service on key accounts destroys more value than it saves. Run sensitivity analysis on SLA tightening before presenting to the CFO.

Key Takeaway: The LP model finds the mathematical optimum, but the implementable optimum is always a constrained version of it. The value of the model is not the final answer it's the structured conversation it forces between logistics, finance, and commercial about what constraints actually matter.